"The are three kinds of lies: lies, damned lies, and statistics."The above quotation, popularised by Mark Twain but probably wrongly-attributed to Benjamin Disraeli is generally trotted out to attack weak positions supported by questionable statistics, or to attack strong positions by disparaging reliance upon statistical information.
However, today we are fascinated by an example of solid statistical analysis used to investigate a plausible proposition, that in fact turns out to be completely without basis. This is a story about what happens when you start out by asking the wrong question, even if everything you do from then on is right! The truth is that statistics do not always set out to be bad, sometimes they're just drawn that way.
APPLICANT DELAY AND THE USPTO BACKLOG
There are presently around 735,000 US patent applications awaiting examination (reported here), while over 8,500 new applications are filed each week (reported here).
Last Thursday on his popular Patently-O blog (well, we like it a lot), Dennis Crouch reported on a question asked by former USPTO Deputy Director Stephen Pinkos at a recent meeting of the USPTO's Patent Public Advisory Committee (PPAC). Specifically, Mr Pinkos asked USPTO officials to identify the portion of the backlog that can be attributed to patent applicant delays rather than to the USPTO.
Now, on the face of it this seems like a reasonable proposition. The PTO is like a "patent pipeline": applications go in at one end, and disposals (issues or abandonments) come out the other. We all know that the volume of a pipeline is proportional to its length, and so the greater the backlog of applications the longer the delay in the system, ie average pendency of applications. This is completely true, and Mr Pinkos' question is based on the complementary proposition, that a longer average pendency time corresponds with a greater backlog. And since applicant delays in responding to Office Actions increase the time for which applications are pending, applicants must share some degree of responsibility for the backlog.
To try to find an answer to Mr Pinkos' question, Dennis Crouch analysed 1,860 randomly selected patents issued between April and July 2010, to determine the delay in each case that was due to applicants responding to Office communications later than the ordinary two and three month deadlines. The result was that the average applicant-induced delay, by this measure, was around two months.
With new applications arriving at a rate of 8,500 per week, this might lead to the conclusion that there are about 70,000 more applications waiting in line than there would be if all applicants responded within the set deadlines, ie nearly 10% of the total backlog!
APPLICANT DELAY DOES NOT CONTRIBUTE TO BACKLOG
Regrettably, however, the entire exercise discussed above is based upon an ill-posed question. Applicant delays do not and cannot contribute to the backlog (although other applicant behaviours may do so, as discussed below).
There are no doubt many ways of demonstrating this fact. The hard way, which Patentology characteristically pursued in a comment in replay to the Patently-O posting, is based upon a simple queueing model in order to describe the process within the PTO. A slightly refined version of this model is described in an annexure to this posting (click here to view, opens in new window). We describe the model by analogy to a queue of customers in a bank, so it should be reasonably accessible to everybody.
An intuitively appealing explanation was offered by contributor "DC". Simply put, while one applicant is failing to engage, the Examiner is filling his or her time examining applications for other applicants. So no other application is delayed, and all other applications keep flowing through the pipeline while the non-responsive applicant effectively stands aside.
APPLICANT BEHAVIOURS THAT ADD TO BACKLOG
We think that the extra effort involved in describing a model of the backlog was worthwhile, because aside from showing that applicant delays do not contribute to backlog, it also clearly identifies in mathematical terms the two changes which between them would have the single greatest impact (other than just hiring a lot more examiners) upon reducing the backlog.
And those changes are:
- limiting the ability of applicants to file continuations and requests for continued examination (RCE's); and
- constraining the time required to perform examination of an application to a given level of quality, which is most readily achieved by limiting the number of claims, and especially independent claims, that an applicant may submit for examination.
In view of the above analysis, it is hardly surprising that the USPTO is pursuing more creative approaches, such as:
- encouraging examiners to welcome, or even initiate, interviews in order to reach agreement on allowable claims earlier in the process, in the hope of reducing the rate of continuations and RCE's;
- providing incentives for applicants to voluntarily withdraw applications from the queue;
- cooperating with foreign patent offices to reduce the search and examination burden on the USPTO; and
- proposing a multi-track examination process, which will enable applicants to "step aside" into the slow track if they do not require earlier patent issue.
The USPTO is not the only national patent office with a long backlog of applications, however it faces a number of unique challenges in addressing the problem. New Director David Kappos appears to be adopting a more consultative and creative approach than that of his predecessor, which is welcome. We will continue to watch developments with interest.